-
Reconstruction of missing features in data using regression (2014)
We often encounter situations where no data value is stored for a particular variable in an observation. A familiar example in particle physics is the decay of the W boson to an electron and a neutrino, where the latter escapes largely undetected. I developed a regression analysis to reconstruct the missing features.
-
Quantitative evaluation of the impact of scientific publications (2014)
Wrote a Python-based tool to query and analyze the impact factor of particle physics publications.
-
Analysis of data from social network (2014)
Social networks constitute a huge amount of consumer data. To take a close look at the kinds of information we can retrieve about people in our network from the social media APIs, I wrote a simple tool to perform statistical analysis of my LinkedIn contacts.
-
Trigger emulator for the Large Hadron Collider (2013−2014)
Developed a C++/Python-based software to emulate features of the next generation of event triggers at the Large Hadron Collider.
-
Pattern recognition to improve signal-over-noise (2012−2013)
Implemented pattern recognition algorithms to detect interesting substructures inside a cluster and to minimize contribution from the ambient noise.
-
Future stock price prediction (2012)
Implemented a lightweight Python code to predict future stock price based on the stochastic nature of the price movement simulated using geometric Brownian motion.
-
Search for the Higgs boson (2011−2013)
Higgs boson gives mass to all fundamental particles in the universe. I developed a series of classification and regression steps to dig out signals of the Higgs boson and/or other novel phenomena from data that also contain a trillion times larger background.
-
Anomaly detection in the first data from the Large Hadron Collider (2009−2010)
Identified a set of features and developed algorithms to detect anomalous events in the early LHC data. A small fraction of events (~0.0001%) that got tagged as anomalous was examined manually for further analysis. Almost all such events turned out to be due to poor measurement. This helped in quickly catching the problems and calibrating the detector.
-
Algorithms to compute efficiency and false-positive (2008−2010)
Developed algorithms to compute event reconstruction efficiency and false-positive in a particle physics experiment. These have been used in hundreds of measurements including the Higgs boson discovery.
-
Clustering and feature extraction (2008−2010)
Developed algorithms to identify particle clusters in an event and calibrate their most important features.
-
Signal detection using machine learning (2006−2007)
Using machine learning developed optimal particle identification criteria, which resulted in a factor of two improvement in the false-positive and gave my collaboration significant advantages over the competitors.
-
Operations of Cherenkov detector (2005−2006)
Ran operations of the Cherenkov detector for my experiment and led its upgrade task, which included replacement of electronics and cooling systems. Have first-hand experience with modern detector control systems, front end electronics, and data acquisition.
- Optimization of signal detection efficiency vs false-positive (2004−2005)
Developed regression models to compute particle detection efficiency in data and simulation. Devised algorithms to tweak simulation to match real data at the sub-percent level. Performed optimizations to minimize false-positive while keeping efficiency as high as possible.
- Biosensor to detect toxic chemicals in water supply (2001−2003)
Worked in a cross-disciplinary team of scientists to develop a low-cost biosensor for continuous monitoring of water supplies for the presence of toxins such as mercury, arsenic, and cadmium. In particular, I focused on the optical signal detection and accurate calibration of the signal.
- Probe of atomic surface using scanning tunneling microscope (2000−2001)
A sharp metal tip placed close to a conductive material in presence of an electric field produces quantum tunneling. I used this tunneling current to probe the sub-nanometer scale topography of a few atoms thick graphite surface. I measured the angles and distances between the atomic rows in 2D and 3D planes.
- Properties of hexagonal ice (2000)
Virtually all ice in the biosphere is in the form of hexagonal crystal. I performed experiments to study its many peculiar properties that are relevant to the existence of life and regulation of global climate.
- Simulation of traffic jam (1999)
Traffic phenomena such as the transition from free to congested flow can be accurately reproduced using cellular automata models. I performed simulation studies aimed to reproduce and solve real life traffic jam problems in the simple case of two-lane street intersections.